sjtu-automatic-maritime-system/onboard-data-collector

LiDAR and camera data collection system, for onboard computer.

Data Collection System for Marine Engineering

Main contributor: @PENG Zhenghao

Setup

PyCapture2

Please download and install PyCapture2 from PointGrey.

Configure your LiDAR

1. Power up your LiDAR
2. Connect it to your computer
3. Close your internet connection
4. Assign a fixed IP for your LiDAR (anyone should be OK)
5. Run the program!

Get Start!

Example usage:

1. Run the data collector for nearly half hour:

    python collect_data.py --exp-name 0101-Trimaran-Tracking --timestep 18000

or

    python collect_data.py --exp-name 0101-Trimaran-Tracking -t 18000

2. Run the data collector for not pre-defined time duration:

    python collect_data.py --exp-name 0101-Trimaran-Tracking

(Note that you should press Ctrl+C to terminate this program!)

3. Load the stored data:

# This is a python script
from recorder import Recorder

config = {"exp_name": "2019-04-16_17-02-20", "save_dir": "experiment"}
r = Recorder(config)
data = r.read()
lidar_data = data["lidar_data"]
frames = data["frame"]
extra_data = data["extra_data"]

print("lidar_data contains {} and its shape is {}.".format(lidar_data, lidar_data.shape))

import cv2

for i in frames:
    cv2.imshow("example", i)
    cv2.waitKey(50)
cv2.destroyAllWindows()

Data Structure

experiment
  +exp-name.h5  # The following is the structure of .h5 file.
    +/config (attrubite)
    +/lidar_data (dataset, a np.ndarray with shape (-1, 30600), is the raw data collected from LiDar)
    +/extra_data (dataset, a np.ndarray with shape (-1, 8), the GPS information)
    +/frame (dataset, a np.ndarray with shape (-1, 960, 1280, 3), camera captured image at the same frequence of LiDAR)
    +/timestamp (dataset, a np.ndarray with shape (-1,))

Note that in current implementation, all data are synced sampled, which means they have same length in temporal dimension.